Monitoring of response to pre-operative chemoradiation in combination with hyperthermia in oesophageal cancer by FDG-PET

Abstract

Purpose: To evaluate the use of positron emission tomography using ¹⁸F-fluorodeoxyglucose (FDG-PET) to assess early response to pre-operative chemoradiation therapy in combination with external locoregional hyperthermia in patients with oesophageal cancer by correlating the reduction of metabolic activity with histopathologic response. Material and methods: Twenty-six patients with histopathologically proven intra-thoracic oesophageal cancer (with ≤2 cm gastric involvement), scheduled to undergo a 5-week course of pre-operative chemoradiation therapy and hyperthermia, were included. FDG-PET was performed before (n = 26) and 2 weeks after initiation of therapy (n = 17). FDG uptake was quantitatively assessed by standardized uptake values. Results: After neoadjuvant therapy, 24 of the 26 patients underwent surgery. In 16 patients changes in FDG uptake were correlated to histopathologic response. In these patients, histopathologic evaluation revealed less than 10% viable tumour cells in eight patients (responders) and more than 10% viable tumour cells in eight patients (non-responders). In responders, FDG uptake decreased by a median −44% (−75 to 2); in non-responders, it decreased by a median of −15% (−46 to 40). At a threshold of 31% decrease of FDG uptake compared with baseline, sensitivity to detect response was 75%, with a corresponding specificity of 75%. The positive and negative predictive values were both 75%. Conclusion: FDG-PET is a promising tool for early response monitoring in patients undergoing chemoradiation therapy in combination with hyperthermia.

• Oesophageal carcinoma
• neoadjuvant thermochemoradiation therapy
• response-monitoring
• FDG-PET

Introduction

The prognosis of patients with locally advanced oesophageal cancer is poor when treated with standard surgery, with overall 5 year survival rates ranging from 20–40% [1], [2]. Many institutes apply neoadjuvant chemoradiotherapy to improve long-term outcome, especially after the recent publication of favourable long-term results of a randomized trial from the Medical Research Council Oesophageal Cancer Working Party, comparing neoadjuvant chemotherapy followed by surgery vs. surgery alone [3]. However, in a large proportion of patients insufficient objective response is achieved. At present, considerable attempt is made to improve the efficacy of pre-operative therapy by using new chemotherapeutic drugs, optimizing the dose of radiotherapy or adding hyperthermia. Promising results have been achieved with the combination of heat and chemotherapy or radiotherapy [4], [5]. Hyperthermia (HT) is a treatment modality defined as an artificial elevation of tissue temperature (40–44°C) for 30–60 min in order to destroy tumour cells. The rationale for thermochemoradiation therapy is bipartite. First, the assumption that heat increases tumour-cell kill by direct cytotoxicity and chemo-sensitization by interference with the repair of DNA damage and by a synergistic interaction with cytotoxic drugs [6]. Secondly, HT selectively kills cells in hypoxic conditions and cells in the S-phase of the cell cycle, which are radioresistent factors.

There is only one randomized study, which reports on HT in combination with chemoradiation in oesophageal carcinoma. In this study, the patients who underwent intra-luminal HT combined with chemoradiation showed an improved 3-year survival rate as compared to pre-operative chemoradiation alone [5]. In the Academic Medical Centre (AMC) an external heating system was developed using the AMC 70 Mhz four antenna array for regional HT [7]. A phase I clinical study has shown that a combination of chemotherapy and external locoregional HT is feasible, together with an acceptable patient tolerance and toxicity [4].

However, despite these important attempts to improve tumour response, there remains a proportion of patients who do not achieve a sufficient objective response but do suffer from toxic side effects. Moreover, prolonged but ineffective pre-operative treatment will inevitably delay appropriate surgical therapy. This down-side of neoadjuvant therapy underlines the need for early identification of non-responders.

Currently, there is no universally accepted, reproducible and reliable means of monitoring response of oesophageal cancer to neoadjuvant therapy.

Conventional techniques for monitoring therapeutic effects in oncology, such as computer tomography (CT) and endoscopic ultrasonography (EUS), are based on morphological imaging [8–10]. General restrictions of these methods include difficulty in distinguishing viable tumour from necrotic or fibrotic tissue and delay between cell kill and tumour shrinkage [11–13]. FDG-PET (positron emission tomography using 18F-2-fluoro-2-deoxy-D-glucose) has been shown sensitive in several malignancies (e.g. breast cancer, lung cancer) for detection of tumour response to therapy. In a systematic review, comparing diagnostic accuracy of FDG-PET in assessing response to neoadjuvant therapy in patients with oesophageal cancer to conventional techniques such as CT and EUS, FDG-PET seemed a promising non-invasive tool [14].

In a more recent study, the role of FDG-PET in distinguishing responders from non-responders early during the course of chemoradiotherapy was investigated [15]. Patients with squamous cell carcinomas received 4 weeks of radiotherapy (2.0 Gy five times a week), concurrently with intravenous infusion of fluorouracil (at a daily dose of 300 mg m⁻²). Although increased FDG uptake caused by radiation induced inflammatory changes in the oesophagus is a well known phenomenon [16–18], Wieder et al. [15] found a decrease of FDG uptake in all patients during chemoradiation, suggesting that inflammatory stress reactions induced by chemoradiotherapy seem to be less relevant for tumour FDG uptake than the therapy induced reduction of viable tumour cells. By applying a cut-off value of 30% the highest accuracy was achieved for differentiation between pathological responders and non-responders, providing a sensitivity of 93% and a specificity of 88%, respectively. However, when combining chemoradiation with additional hyperthermia, the cut-off value is not automatically the same as proposed by Wieder et al. [15]. The additive effect of HT on changes in tumour FDG uptake is as yet unclear. The possible contribution of HT to local inflammation [19] might result in enhanced FDG uptake, thus making interpretation of early response monitoring with PET more problematic. Data evaluating early response to thermochemoradiation (ThCR) are lacking.

Therefore, the aim of this study was to assess the accuracy of FDG-PET in early response monitoring during chemoradiation in combination with external locoregional hyperthermia.

Patients and methods

Patients

Between August 2003–January 2005, 26 consecutive patients with potentially curable histologically proven carcinoma of the oesophagus with ≤2 cm gastric involvement consented to participate in the present prospective non-randomized study. Exclusion criteria were diabetes; uncontrolled heart failure, hypertension, severe arrythmia and pacemaker; pre-existing myelopathy and/or polyneuropathy; previous radiotherapy and/or chemotherapy. All patients were staged with multi-detector computed tomography (MDCT), endoscopic ultrasound (EUS) and external ultrasound of the neck, both combined with fine needle aspiration (FNA) on indication. Patients without evidence of distant metastases and/or locally irresectable disease on this conventional pre-treatment work-up underwent a baseline FDG-PET.

After these diagnostic procedures, patients underwent concurrent chemotherapy, radiotherapy and hyperthermia for 5 weeks. Assessment of tumour response by FDG-PET was performed 2 weeks after initiation of neoadjuvant therapy. Patients without signs of tumour progression or other contraindications for oesophagectomy were scheduled to undergo surgery. The study protocol was designed to allow comparison of FDG-PET data and histopathologic examination of tumour tissue. It was approved by the medical ethical committee at the Academic Medical Centre Amsterdam. Written informed consent was obtained from all patients.

Neoadjuvant treatment

Neoadjuvant therapy consisted of a 5 week schedule of chemoradiation and hyperthermia. The schedule was as follows: chemotherapy, consisting of paclitaxel 50 mg m⁻² and carboplatin AUC = 2, by intravenous infusion on days 1, 8, 15, 22 and 29; radiotherapy consisting of a total of 41.4 Gy, given in 23 fractions of 1.8 Gy, five fractions per week, starting the first day of the first cycle of chemotherapy; and hyperthermia with concurrent chemotherapy administration within 1 h after the radiotherapy treatment. The hyperthermia method was described earlier by Albregts et al. [4]. Patients were treated in a prone position with the four waveguide AMC array operating coherently at 70 MHz from four sites around the thorax. A power maximum is created at the tumour, using an intra-oesophageal E-field probe to obtain optimal phase settings. The e-field probe was positioned endoscopically at tumour level before each hyperthermia treatment. Total power ranged between 800–1000 W, distributed over the waveguides in a ratio of 1:3:3:3 for the top, bottom, left and right waveguide, respectively, to prevent neural toxicity in the spinal cord. Skin toxicity was prevented by circulating cold water maintained at 13°C in boli on the skin. The temperature in the oesophagus, rectum and intra-muscularly near the spine were monitored every 30 s using 14 and 21 sensor thermocouple probes. The temperature and thermal dose parameters T₁₀, T₅₀, T₉₀ were derived from the tumour temperature data.

Four-to-six weeks after neoadjuvant therapy, patients were scheduled for potentially curative oesophagectomy by either a transhiatal or transthoracic approach.

PET imaging

A baseline FDG-PET was performed during initial staging before initiation of pre-operative therapy. FDG-PET was repeated on day 15, 6 days after the second cycle of chemotherapy and hyperthermia, and 2 days after the tenth fraction of radiotherapy. FDG-PET scans were performed using an ECAT EXACT HR + PET scanner (Siemens/CTI Inc., Knoxville, TN). Prior to PET imaging, patients were instructed to fast for at least 6 h. Patients were also instructed to drink 1000 mL of water prior to imaging to stimulate FDG excretion from the renal calyces and to stimulate subsequent voiding. Blood glucose levels were measured before and after each PET examination.

Scans were performed in a 2D acquisition mode, using an interleave emission-transmission scanning protocol. Emission scans (first scan from mid-skull to mid-femur, follow-up scan only from the primary tumour) of 5 min per position were performed 90 min after intravenous injection of 250–370 MBq FDG. Transmission images were obtained for 3 min per bed position for attenuation correction. All scans were corrected for decay, scatter and randoms and reconstructed using ordered sub-set expectation maximization (OSEM) with two iterations and 16 sub-sets followed by post-smoothing of the reconstructed image using a 5 mm at full width at half maximum (FWHM) Gaussian filter.

Evaluation of FDG-PET

For the quantitative evaluation of regional FDG uptake, volume of interest (VOI) was generated using a 3D region-growing algorithm with in-home developed software [20], [21]. To assure that the pre- and post-therapy SUV measurements were taken in the same region of the tumour, post-therapy VOIs (volumes of interest) were always correlated to baseline scanning (scanner settings were identical before and after therapy using internal landmarks (indicated in cm from the jugular notch)). A hotspot of the primary tumour was always clearly defined.

In short, VOIs were generated as follows: first the location of maximum uptake was approximately indicated by the observer. Starting from this pixel a 3D 50% isocount contour was generated including all pixels with a value higher than 50% of the initially indicated pixel. This first VOI roughly indicated the boundary of the tumour in 3D. Within this first VOI, the pixel with the absolute overall maximum pixel value in the tumour was located and a second region growing step was initiated to determine the final 50% isocount 3D volume of interest. The second and final region growing procedure used 50% of the absolute maximum pixel value as threshold. Average SUVs were calculated using the average pixel value within that final 50% VOI.

To assess FDG uptake semi-quantitatively, the standardized uptake value corrected for body surface area and plasma glucose concentration (SUV_BSA–glu) was used. SUV_BSA–glu was calculated according to the equation: SUV_BSA–glu = ((C_t/(I/BSA)) × C_glu), with C_t as the tissue activity concentration, I as the injected dose, BSA body surface area (=0.007184 × weight^0.425 × height^0.725) and C_glu as the plasma concentration of glucose.

Surgical therapy

All patients who were considered suitable for surgery and did not demonstrate tumour progression after neoadjuvant therapy underwent surgical treatment, consisting of transhiatal or transthoracic oesophagectomy followed by gastric tube reconstruction.

Histopathologic evaluation

Resection specimens were fixed with formaldehyde (4%) for 24 h. Processing of the resected specimens was done using a standardized protocol. Tissue samples were routinely taken from the specimens, including proximal and distal resection margins, longitudinal strips through the primary tumour and adjacent mucosa and normal appearing mucosa proximal and distal to the tumour. If no macroscopic tumour was visible, the entire mucosal surface was dissected in longitudinal direction, bound proximally by squamous epithelium and distally by gastric epithelium.

All specimens were evaluated by an experienced gastrointestinal pathologist (FJtK), who was unaware of the clinical and PET data, in accordance with the criteria of the International Union Against Cancer, including stage, R classification and grade. Tumours were classified as histopathologically responding when less than 10% viable tumour tissue was found in the tumour bed; otherwise the tumour was classified as histopathologically non-responding. [15], [22] The group of non-responders was further divided into patients with partial response (10–50% viable tumour tissue), minimal response (>50% viable residual tumour tissue) and no change (absence of any regressive changes). In the group of responders, response was classified as complete (histologic fibrosis with no viable residual tumour cells) and sub-total response (<10% viable residual tumour cells) [23].

Statistical analyses

Statistical analyses were performed using SPSS for Windows 11.0.1 (SPSS, Inc., Chicago, IL). All quantitative data were expressed as median with range (minimum and maximum) and differences between groups were tested by the Mann–Whitney U-test. For intra-individual comparisons before and after pre-operative therapy, a Wilcoxon signed rank test was applied. Differences in FDG uptake before treatment and after 2 weeks of treatment were correlated with pathology of the resected specimen. Receiver operating characteristic (ROC) curves [24] were used to evaluate the diagnostic accuracy of FDG-PET for assessment of histopathologic response. For calculation of these curves, the threshold value for definition of a tumour response in PET imaging was systematically varied over the full range of the observed changes in tumour FDG uptake. For each of these threshold values the percentage of correctly predicted histopathologic responses (sensitivity, true positive rate on the y-axis) was plotted against the rate of incorrectly predicted histopathologic responses (1-specificity, false-positive rate on the x-axis). The optimum threshold value for differentiation of responding and non-responding tumours was defined by the point on the ROC curve with minimum distance from the 0% false-positive rate and the 100% true-positive rate. For this threshold value sensitivity, specificity and positive and negative predictive values were calculated using standard formulae.

Results

Demographic data

The mean age of the 26 patients was 61 (range 29–73). There were two female and 24 male patients. The primary tumour (20 adenocarcinomas and six squamous cell carcinomas) stage on pre-therapeutic endosonography was uT1 in one patient, uT2 in three patients, uT3 in 21 patients and uT4 in one patient.

Surgery

Twenty four of the 26 patients underwent surgical therapy. One patient died after completing neoadjuvant therapy due to pulmonary embolism. One patient received further palliative treatment due to progressive disease during neoadjuvant therapy. Fifteen patients underwent a transhiatal and nine patients a transthoracic oesophagectomy. A complete resection (R0) was achieved in all patients.

Histopathologic analysis and response evaluation

The distribution of the pT and pN categories and the stage grouping are shown in Table I. The histopathologic response evaluation in the resected specimen revealed complete response in five of the 24 patients (21%), sub-total response in five patients (21%), partial response in six patients (25%), minimal response in seven patients (29%) and no change in one patient (4%).

Table I.  Patient characteristics of responders and non-responders, changes in tumour FDG uptake and histopathologic tumour response.

No.	Sex	AC/SCC	Clinical staging (pre-ThCR)	Pathological staging (post-ThCR)	PA response	SUVBSA-glu at baseline	SUVBSA-glu after 2 weeks of ThCR	Change in SUV (%)
Responders
1	M	SCC	T3N1	T0N0	Complete	0.38	–	–
2	F	SCC	T3N0	T0N0	Complete	0.23	0.11	−50
3	M	SCC	T2N1	T0N0	Complete	0.17	0.06	−63
4	M	AC	T3N0	T0N0	Complete	0.23	0.16	−32
5	M	SCC	T3N1	T0N0	Complete	0.40	0.10	−75
6	M	AC	T1N1	T2N1	Subtotal	ND	–	–
7	M	SCC	T3N0	T2N0	Subtotal	0.21	0.14	−30
8	M	AC	T3N1	T3N1	Subtotal	0.50	0.15	−71
9	M	AC	T3N1	T2N0	Subtotal	0.17	0.11	−31
10	M	AC	T3N0	T3N1	Subtotal	0.10	0.11	2
Non-responders
11	M	AC	T3N1	T3N1	Partial	ND	–	–
12	M	AC	T3N1	T1N0	Partial	0.23	–	–
13	M	AC	T3N1	T3N1	Partial	0.11	0.11	3
14	M	AC	T3N1	T1N1	Partial	0.15	0.11	−31
15	M	AC	T3N1	T3N0	Partial	0.22	0.19	−13
16	M	AC	T3N0	T2N0	Partial	0.16	0.15	−2
17	M	AC	T2N0	T1N0	Minimal	ND	–	–
18	M	AC	T3N0	T3N1	Minimal	ND	–	–
19	M	AC	T2N0	T1N0	Minimal	0.32	–	–
20	M	AC	T3N1	T3N0	Minimal	0.13	0.10	−24
21	M	AC	T3N1	T3N1	Minimal	0.32	0.44	40
22	F	SCC	T3N1	T2N1	Minimal	0.46	0.26	−44
23	M	AC	T3N0	T3N1	Minimal	0.22	0.12	−46
24	M	AC	T3N1	T3N1	No change	ND	–	–
25	M	AC	T4N1		NA	0.33	–	–
26	M	AC	T3N1		NA	0.10	0.11	11

No., patient number; AC, adenocarcinoma; SCC, squamous cell carcinoma; ThCR, thermochemoradiation; F, female; M, male; NA, not applicable; SUV_BSA-glu, standardized uptake value; ND, non-detection.

FDG uptake before and after treatment

All patients underwent PET imaging before treatment and 17 patients underwent imaging 2 weeks after initiation of treatment. Nine patients underwent no second imaging during treatment due to the following reasons: no uptake in the primary tumour (n = 5), no fasting (n = 1), technical problems (n = 2) and refusal (n = 1). In one patient who underwent a PET scan before and 2 weeks after initiation of therapy, no histopathology could be obtained, due to fatal pulmonary embolism before surgical treatment and this patient was, therefore, left out of response analysis.

The SUV_BSA–glu averaged to 0.24 (0.10–0.50) before treatment and decreased significantly to 0.15 (0.06–0.44) 2 weeks after initiation of therapy (n = 16; median relative decrease = −31%, range −75 to 40%, p = 0.008). Thirteen of the tumours (81%) showed a decrease of FDG uptake, three tumours showed an increase of 2, 3 and 40% of FDG uptake after two cycles of therapy. The patient with 40% increase of FDG uptake developed an oesophageal abscess that perforated into the lung, after completing ThCR. Individual data are shown in Table I and Figure 1(a) and (b).

Figure 1. Individual standardized uptake value (SUV) data of responders and non-responders, before and after 2 weeks of initiation of the neoadjuvant thermochemoradiation therapy (ThCR).

Correlation of findings in FDG-PET and histopathologic response

Typical examples of FDG-PET scans in histopathologically responding and non-responding patients are depicted in Figure 2.

Figure 2. Examples of fluorodeoxyglucose (FDG) positron emission tomography studies (coronal and axial slices) before and 2 weeks after thermochemoradiation therapy, in (a) histopathologically responding and (b) non-responding tumours.

The absolute SUV_BSA–glu of the tumours at the time of the baseline PET and the PET after 2 weeks of therapy were not significantly different between histopathologic responders and non-responders (Table II). However, there was a significant correlation between the relative decrease in FDG uptake and histopathologic response (Table II). Histopathologically responding tumours (n = 8) showed a decrease in SUV_BSA-glu of median of −44% (−75 to 2%) whereas non-responding tumours (n = 8) showed a decrease of only median of −15% (−46 to 40%, p = 0.05, Table II).

Table II.  Patient characteristics, changes in standardized uptake value (SUV).

	Responders	Non-responders	p-value*
Number of patients	8	8	NS
Age (range)	3 (50–73)	7 (29–73)	NS
Male/Female	7:1	7:1	NS
Histology (AC:SCC)	4:4	7:1	NS
SUV at baseline	0.25 (0.10–0.50)	0.22 (0.11–0.46)	NS
SUV after 2 weeks ThCR	0.11 (0.06–0.16)	0.18 (0.10–0.44)	NS
Median decrease SUV	−44% (−75 to 2%)	−15 (−46 to 40)	0.05

AC, adenocarcinoma; SCC, squamous cell carcinoma; SUV, standardized uptake value; ThCR, thermochemoradiation; NS, not significant.

*Mann–Whitney test.

The ROC curve demonstrated the highest accuracy for differentiation between responding and non-responding tumours at a cut-off value of a 31% decrease of baseline FDG uptake (Figure 3). Using this cut-off value, six of the eight responding and six of the eight non-responding tumours were correctly identified, providing a sensitivity of 75% (95% CI 41–93) and a specificity of 75% (95% CI 41–93). Positive and negative predictive values were 75 and 75%, respectively.

Figure 3. Receiver operating characteristic curve for assessment of histopathologic response by fluorodeoxyglucose (FDG) positron emission tomography. Changes in FDG uptake from the baseline scan to the scan after 14 days of thermochemoradiation therapy.

Clinical outcome

Survival at a median follow-up of 9 months (range 5–20 months) was 91%. All responders were still alive without evidence of disease at the end of follow-up.

Correlation of findings in FDG-PET and temperature

Figure 4 shows the correlation between four ascending groups of SUV decrease (in each group four patients) and the average temperature of all hyperthermia sessions. The correlations between SUV decrease and minimum or maximum tumour temperatures were similar. Average temperature of patients with a SUV decrease of < or > 31% (eight patients in both groups) was 38.8 and 39.7°C, respectively. Although there was a trend to increased SUV decrease with higher temperature, this was not statistically significant, mainly due to the small numbers.

Figure 4. Average temperature vs. SUV decrease. Each dot represents the mean SUV decrease in four patients with ascending SUV decrease values.

Discussion

This pilot study shows that FDG-PET is a promising tool for early response monitoring, in patients undergoing pre-operative chemoradiation therapy in combination with external locoregional hyperthermia. There was a significant difference in FDG uptake decrease between responders and non-responders to pre-operative therapy.

A few studies analysed the value of FDG-PET to assess early tumour response during neoadjuvant therapy in oesophageal carcinoma [15], [25], [26]. However, patients in these earlier studies were pre-operatively treated with chemotherapy alone [25], [26] or a combination of chemotherapy and radiotherapy [15]. To the authors’ knowledge, this is the first study, investigating the value of FDG-PET for early response monitoring in oesophageal cancer patients receiving chemoradiation in combination with hyperthermia. As mentioned in the introduction, the additive effect of hyperthermia on the PET signal is still unknown, but theoretically it is possible that HT, if combined with chemoradiation therapy, amplifies changes in glucose metabolism in normal surrounding tissue and tumour. However, the accuracy of FDG-PET in this study was not significantly different from that of the study of Wieder et al. [15] (sensitivity 75% (95% CI 41–93%) vs. 93% (95% CI 68–100%), specificity 75% (95% CI 41–93%) vs. 88% (95% CI 47–100%), respectively. Although Wieder et al. [15] used another chemoradiation scheme, response rates were comparable (Wieder et al. 57% vs. 46% in the present study) and responders and non-responders showed similar FDG decrease in both studies (responders 44 and 44%, respectively, non-responders 15% and 21%, respectively). Moreover, the addition of hyperthermia did not lead to an evidently different ROC curve. Therefore, the possible confounding effects induced by hyperthermia on the PET signal seem to be minor and not to hamper discrimination between responders and non-responders. This is in agreement with comparable studies evaluating response to multi-modality therapy including hyperthermia in patients with rectal cancer [27], [28]. With the relevant remark, that these studies measured response 2–4 weeks after full completion of pre-operative therapy, instead of early during therapy.

The 31% cut-off value for a metabolic response was chosen to ensure the highest accuracy, considering false-positivity as important as false-negativity from a clinical point of view. However, the results of a phase II trial currently performed in the AMC, especially the balance between therapeutic effectiveness and side-effects of ThCR, should ultimately determine the optimal cut-off value. For example, if the results from this new phase II trial indicate that the induced side-effects are low and responders have a significantly better survival than non-responders, the test should have a sufficiently high negative predictive value to be useful in clinical practice. In clinical practice, a negative test (i.e. no response on PET) would imply that ThCR therapy would be discontinued. If the negative predictive value of PET is not sufficiently high, this would too often lead to erroneous discontinuation of ThCR in responders, incorrectly classified by PET as being non-responders.

Addition of hyperthermia to chemoradiation therapy could be a promising effort to improve prognosis, however its efficacy is yet unknown. Particularly in patients who undergo an intensive, possibly contra-effective therapy (inevitably delaying surgical therapy), early identification of non-responders is of great importance. In this study, changes in FDG uptake correlated well with pathologic response in patients undergoing thermochemoradiation. Therefore, it is concluded that FDG-PET is a promising tool in early response monitoring in these patients, although its exact role should be determined in larger trials.